Understanding Cancer Outcomes: Breast vs. Cervical
About the Project
Our project investigates survival patterns in breast cancer and cervical cancer using data from The Cancer Genome Atlas. Cancer patients with the same diagnosis expirence different outcomes based on several factors whether it is treatment recieved, age, cancer stage and other clinical factors. Understanding these patterns will assist in improving patient care and personalizing medicines. Our analysis combines unsupervised learning and supervised learning to identify patient subgroups, predict survival time, and identify top important prognotic factors.
Key Questions: - What distinct patient subgroups exist within breast and cervical cancer? - How do mortality patterns differ between these cancers, and why? - Which clinical features most strongly predict survival outcomes? - What are the key prognostic factors for each cancer type?
Literature Review
To better understand survival patterns in breast cancer (BRCA) and cervical cancer (CESC), multiple intersecting factors should be investigated. The Cancer Genome Atlas (TCGA) has revealed that even with the same diagnosis, different cancer subtypes exist. Known clinical features that predict survival include stage at diagnosis, age, and treatment type, but these may not be equally important for cervical versus breast cancer.
Research on cervical cancer discovered patterns that predict treatment response while, breast cancer research has categorized patients based on hormone receptors and genetic characteristics. These cancers also differ in their key predictive factors: cervical cancer outcomes are heavily influenced by HPV status, cell type, and early detection via pap smears, while breast cancer outcomes are influenced by hormone receptor status and HER2 amplification. Treatment approaches also differ substantially, with breast cancer benefiting more from targeted therapies and cervical cancer relying more on radiation and surgery.
Given these complex differences, machine learning offers powerful analytical tools. Unsupervised learning methods have shown to identify patient subgroups by revealing cluster structures in clinical and demographic data. Random Forest models excel at predicting survival and comparing feature importance across cancers. Our study combines both unsupervised and supervised approaches, providing a comprehensive framework for understanding what determines survival in cancer patients.
About Us
::: column width=“50%” 
{width=70% alt = “Munashe Mhlanga portrait”}
Munashe Mhlanga
MS Data Science & Analytics, Georgetown University
Munashe focuses on applied data science for social impact, with research interests in digital transformation, accessibility, and ethical data use. LinkedIn
::: column width=“50%” 
{width=70% alt = “Betsy portrait”}
Betsiet Delelegn
MS Data Science & Analytics, Georgetown University
LinkedIn ::: :::
Additional Resources
Please find additional resources on womens health below:
Podcast: Women’s Health
Women’s Health Podcast
A wide-ranging podcast covering women’s health topics including hormonal health, aging, disease prevention, and overall wellbeing.
YouTube Channel: Cleveland Clinic
Speaking of Women’s Health (Cleveland Clinic)
Expert-led video discussions focused on women’s wellness, preventive care, and clinical health topics.
Blog / Reading Resource
Speaking of Women’s Health – Cleveland Clinic
An evidence-based blog and educational resource with articles on gynecology, cancer screening, menopause, and general women’s health.
Learn More About Care
Georgetown Lombardi Comprehensive Cancer Center – Breast Cancer Care
Clinical information, diagnostic pathways, and patient-centered breast cancer care resources.
Donation
Please consider donating to CancerServe, a voluntary organization that focuses on education, community mobilisation, and assistance for disadvantaged patients in Zimbabwe.